Mistral Large 4

Strengths

• Mistral says it is the largest announced open model from a European company, at about a trillion parameters. It was trained on 3,800 NVIDIA Grace Blackwell chips in Mistral's European data centres.
• A 1 million token context window in the API, so it can take in very long documents at once.
• Strong at cybersecurity: Mistral reports 93% on Cybench and a top five place on the Artificial Analysis Cyber Index.
• Trained on more than 160 languages, including every official EU language such as Slovak and Czech.
• Output is cheaper than on the smaller Mistral Medium 3.5: $4.18 per million tokens versus $7.50.

Weaknesses

• It is still a preview. Mistral is refining the model and has not published architecture details or how it was post-trained.
• Neither the weights nor the license were published as of 28 October 2026, the Hugging Face page only announces them.
• In a blind coding evaluation by Surge AI that Mistral itself cites, it scored 3.74 out of 5, while Claude Opus 5 scored 4.22.
• Most of the test results were published by Mistral itself.
• Huge for self-hosting: a trillion parameters needs data centre hardware, not an ordinary company server.

Best for

• demanding tasks with data kept in the EU
• security analysis
• agents that handle several steps on their own
• long documents up to 1 million tokens
• running a strong open model on your own servers later on

Not ideal for

• production use that needs a stable final release
• small company servers

How to use it in Europe

For now you use Large 4 through the Mistral API or Mistral Studio as a public preview. It is available in the EU, and data is processed in the EU by default. Mistral has not said whether it is available in the Vibe app (formerly Le Chat). The weights are due on Hugging Face but had not been published as of 28 October 2026.

What is coming

Confirmed by vendor Mistral promised to publish the weights of its largest model by the end of October. Its official Hugging Face page lists 29 October 2026 as the expected date. When: 29 October 2026, the estimate on Mistral's official Hugging Face page →

Price compared with similar models

Benchmark results

Benchmarks cover only a narrow slice of ability and vendors often report them at their best settings. Treat them as a rough guide.

More models from this vendor

• Mistral Medium 3.5: Mistral's mid-size open-weight model and the default in the Vibe app. Good for everyday work and coding with data kept in the EU.
• Mistral Small 4: Mistral's cheap, fast open model under Apache 2.0 for everyday tasks through the API or on your own server.
• Ministral 3 14B: Mistral's small open model under Apache 2.0 that runs on a single GPU, even without an internet connection.
• Codestral 25.08: Mistral's model for code completion right in your editor. Fast and cheap, but dated for autonomous coding agents.
• Mistral Large 3: Mistral's previous large model with open weights under Apache 2.0. Cheap in the API and suited to your own powerful server.

Similar models from other vendors

• Claude Opus 5.5 (Anthropic): Anthropic's main model for demanding work, coding and long documents, and the one Anthropic recommends starting with.
• DeepSeek V4 Pro (DeepSeek): DeepSeek's largest open-weight model for demanding text tasks at a low price. It does not understand images.
• Gemini 4 Carbon (Google): Google's most capable model, announced on 28 October 2026. So far only selected cybersecurity experts have access.
• Gemini 4 Argon (Google): Google's most capable publicly available model, out in October 2026. Built for demanding analytical work where accuracy matters and made-up answers are costly.
• Muse Wonder (Meta): Meta's strongest model for long-running agents and coding, still released as a preview.
• Kimi K3.1 Pro (Moonshot AI): Moonshot AI's newest model from October 2026, built for agentic coding, office documents and very long outputs.

• AI with EU data: which models suit companies under GDPR
• Mistral Vibe (Le Chat) pricing: what Pro and Team cost
• All AI prices in one place

Sources

• Mistral: Introducing Mistral Large 4 ↗
• Mistral docs: Mistral Large 4 ↗
• Hugging Face: mistralai/Mistral-Large-4-1T-A52B ↗
• Mistral docs: Changelog ↗
• Mistral help: Where do you store my data ↗